Summary
GenBank and GISAID serve as the primary global repositories for viral genomic data, yet integrating records across them is heavily hindered by a lack of standardized cross-references. Because researchers frequently upload the same biological sample to both databases independently—often to satisfy rapid public health sharing alongside formal publication requirements—combined datasets suffer from widespread, unflagged "double-counting." At the same time, rapid growth in decentralized sequencing has increased the prevalence of sequencing and assembly technical artifacts, which artificially inflate substitution rates and degrade temporal signals in evolutionary analyses.
This study introduces a formal, reproducible curation workflow combining two complementary open-source tools:
G2G matcher: Designed for automated cross-repository harmonization and deduplication between GenBank and GISAID (github.com/andrezaleite/G2G-Matcher).
Lab-Specific Bias FILTer (LSBFILT): Designed for targeted masking of laboratory-specific sequencing and assembly artifacts without discarding entire genomes (github.com/khourious/LSBFILT).
Using the Eastern/Central/South African (ECSA) chikungunya virus (CHIKV) lineage as a proof-of-concept, the workflow successfully restores temporal signals for molecular clock modeling. Restricting homoplastic site-masking to problematic sequences reduced the estimated substitution rate from an inflated 8.517 × 10−4 to 5.078 × 10−4 substitutions/site/year and nearly doubled the coefficient of determination (R2) in root-to-tip regression analyses from 0.353 to 0.677, providing a robust dataset for downstream phylodynamic inference.
